Validation of IR Sounder Temperature Profiles Using GPS Radio Occultation
Michelle Feltz, Robert Knuteson, Steve Ackerman, Eva Borbas,
Dave Tobin, and Henry Revercomb
Cooperative Institute for Meteorological Satellite Studies, University of WisconsinMadison
Abstract
This paper presents a validation of the temperature profile retrievals from the advanced infrared and
microwave sounders through inter-comparison with GPS RO profiles matched in time and space. We will discuss
the unique issues related to matchups in time and space between GPS RO and polar orbiting satellites. In
particular we show the impact of including the horizontal averaging inherent in the radio occultation
measurements on the statistical validation. In particular, an analysis of the NASA AIRS Level 2 products (v5) will be
presented compared to measurements made by the COSMIC program. Systematic bias and RMS vertical profiles
have been created for 1 km vertical layers in several latitude zones (Arctic, NH Mid-Lat, Tropical, SH Mid-Lat, and
Antarctic). Application of this method to JPSS CrIMSS data is in progress.
COSMIC AIRS Matchup Methodology
Satellite Data
The COSMIC Data Analysis and Archival CenterCDAAC was used to obtain COSMIC data.
http://cosmicio.cosmic.ucar.edu/cdaac/products.html
The
product used was COSMIC version 2010.2640. A
typical COSMIC profile is obtained in about 100
seconds with over 3,000 vertical samples. The
netcdf files contain time in units of GPS seconds,
which are 15 seconds ahead of UTC. The netcdf
files also contain azimuth angle of the occultation
plane at tangent point with respect North
direction, positive to the East from the North
direction. The angle is measured between North
and the GPS direction of the ray path. A quality
control flag is included in the GPS RO netcdf files.
For an example day, 19 October 2007, the
percentage of GPS profiles marked bad was 2.5%.
These bad profiles are excluded from the analysis.
Figure. Locations of COSMIC RO profiles (blue dots) and AIRS granule bounding boxes (black
squares) with COSMIC RO matchups within one hour (red stars) for 19 October 2007.
http://www.cosmic.ucar.edu/launch/GPS_RO_cartoon.jpg
Following Yunck et al. 2009, we computed the bias and RMS statistics for Arctic, Northern
Mid-Latitude, Tropical, Southern Mid-Latitude, and Antarctic zones. Figures 1 and 2 show
the results for these latitude zones on the CrIMSS Focus Day 19 October 2007 along with
global statistics. Figure 1 illustrates the bias and RMS profiles on the AIRS 100 pressure levels
while Figure 2 contain the same matchup cases degraded to 1 km vertical resolution. Shown
on each figure are both the statistics of the closest profile to AIRS (solid line) and the
spatially weighted average AIRS profile (dashed line).
Focus Day Bias and RMS
AIRS 100 Levels
60
19 October 2007
1 – km Layers
Global
0
Arctic
(90N – 60N)
COSMIC RO profile showing horizontal
ray path extent of 150-300 km
AIRS looks toward the ground through a
cross-track rotary scan mirror which provides
+/- 49.5 degrees (from nadir) ground
coverage along with views to cold space and
to on-board spectral and radiometric
calibration sources every scan cycle. The
scan cycle repeats every 8/3 seconds. Ninety
ground footprints are observed each scan.
One spectrum with all 2378 spectral samples
is obtained for each footprint. A ground
footprint every 22.4 ms. The AIRS IR spatial
resolution is 13.5 km at nadir from the 705.3
km orbit. The AIRS Level 2 sounding products
contain atmospheric temperature and
moisture profiles from about 1 mb down to
the surface with a horizontal resolution of
about 45 km.
As a preliminary approach in our matchup
analysis, the AIRS fields of view within a
circle of radius 150 km centered at the
latitude/longitude of the perigee of the
COSMIC RO profile were found. To
investigate the effect of the horizontal
resolution of the COSMIC RO profiles, the
average AIRS profile within 150 km was
compared with the closest AIRS profile.
Figure 6 shows a comparison of the closest
AIRS profile to the area average AIRS profile
and to a coincident COSMIC RO profile.
North
Mid-Latitudes
(60N – 30N)
Tropics
(30N - 30S)
Figure. COSMIC RO profile (red line) with
horizontal resolution (green lines) overlaid
on AIRS fields of view (black dots).
South
Mid-Latitudes
(30S – 60S)
Antarctica
(60S 90S)
Table 1. COSMIC/AIRS Matchup Yield
Figure 1. AIRS 100 pressure levels
Figure 2. 1 km layers
Conclusions
We developed the matchup tools for the comparison of COSMIC RO and AIRS IR temperature profiles. We
found that one hour time difference between GPS RO and the IR profile measurement provides adequate
yield with sufficient accuracy to characterize bias and RMS errors for 20 degree latitude zones. When GPS RO
horizontal averaging was applied to the AIRS Level 2 granule matchups, improved agreement was obtained
compared to using the closest AIRS IR profile. Daily zonal statistics were found to provide a useful measure of
AIRS retrieval performance. We anticipate that monthly time series for latitude zones can be used to monitor
CrIMSS retrieval performance.
This study shows RMS error is between 1.5 K and about 2 K for the 300 mb to 30 mb altitude range for all
latitude zones. Above this range, the COSMIC RO profiles appear to have higher vertical resolution than the
AIRS IR profiles. Below this range, the COSMIC RO profiles become strongly contaminated by water vapor in
the occultation ray path. The height of this effect is latitude dependent. These results are in qualitative
agreement with results of Yunck et al. 2009.
Future work includes the application of the AIRS temperature averaging kernel to the COSMIC minus AIRS
profile differences to remove vertical structure higher than the theoretical AIRS resolution. Application of
this method to CrIMSS EDR products is in progress and will be reported in a future paper.
Contact: michelle.feltz@ssec.wisc.edu of the University of Wisconsin-Madison AOS/CIMSS/SSEC
Figure. AIRS minus COSMIC RMS statistics versus
latitude for 20 degree latitude zones.
Acknowledgments
We gratefully acknowledge the NPSO (Taiwan’s National Space Organization) and UCAR (University Center for Atmospheric Research) for
access to the COSMIC RO data. We would like to thank the Goddard Space Flight Center (GSFC) Data Archive for access to the AIRS Level 2 data
products. This work was supported under JPSS NOAA grant NA10NES4400013.